IRB protocol review where every required consent element is tied to its paragraph, and people make every determination
Every protocol back to its investigator on the first day, each fix tied to its rule.
See one case, screen by screen ↓IRB Protocol Pre-Review is the staff step of IRB protocol review, done with agents: IRB staff screen protocols, parental permission and consent forms for Common Rule elements before a reviewer sees them. Seven agents check completeness for the review path, tie every required consent element to its paragraph, compare the protocol with every form, read translated forms back, measure reading level, draft a cited return, compare each resubmission with the last version and build a sealed audit pack. The analyst completes the pre-review, a second analyst checks every return, and a designated reviewer, the chair or the board makes every determination.
AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →
Updated
IRB protocol review starts with pre-review. Before a designated reviewer or the board sees a human subjects study, an IRB analyst reads the protocol, the parental permission and consent forms, the child assent script, the surveys, the translated forms and the partner school district's approval, and checks them against the Common Rule (45 CFR 46), Subpart D for children, FERPA and PPRA when student records or surveys are used, and the office's own policies. Most of that is reading and comparing: is each required element there, do the forms say what the protocol says, does the Spanish form mean what the English one means, can a parent read it.
Today the analyst works from an informed consent checklist that says an element is "present", not which sentence meets it, and cross-reads the documents from memory. A permission form that promises two surveys when the protocol schedules three slips through; a translated form is checked for presence, not meaning; the return goes back in comment boxes with no links to the paragraphs, and comes back in rounds. IRB Protocol Pre-Review gives that reading to seven agents, with every statement cited, and leaves every decision with the IRB office.
Estimated split of the demo's 9-day median pre-review (Halden State, 2025–26, fictional), by hand and with the agents; the 'after' column adds up to the 1-day target. Our arithmetic, kept modest.
Seven agents read every submission, tie each consent element to its paragraph and draft the return; the analyst completes it and a reviewer decides.
Protocol 2026-0418, "Paired Reading for Fluency in Grades 3–5", from Dr. Leah Castellano at Halden State University, was submitted at 4:52 p.m. on Tuesday, October 20, for expedited review under category 7 and Subpart D 46.404. It runs in six Marlow Bay Unified elementary schools. Here is what happens to it, screen by screen, in the working solution.
The agents pre-read the study by 5:08 p.m. the day before. Jae-won opens the parental permission form, typeset as a family will read it, with a column of required elements down its right edge. Above it sit the reading level, grade 12.4 against a target of 8, and the one place the protocol and the form disagree.
Surveys: protocol §6.2 says 3, at weeks 1, 9 and 18; the form ¶3 says 2.
Risks, benefits, alternatives, confidentiality, records shared under the FERPA studies exception, voluntary participation and future research each sit beside their paragraph, tied by a thin line. Injury compensation, 46.116(b)(6), is marked does not apply for a minimal-risk study. On the English form, 11 of 14 elements are met, one meets the rule but disagrees with the protocol, and one needs Jae-won.
The agent compares the facts a family or teacher relies on across the protocol, the permission form, the assent script and the teacher consent: paired reading sessions, the Reading Attitude Survey, recordings, district records, retention, compensation and withdrawal. Six agree. The permission form says the survey is given twice; the protocol and the assent script say three times, and a click opens §6.2 on page 9.
The script read to children in grades 3–5 is at reading grade 3.6 against a target of 4, and six of its elements are met. One is not: the key points that should open the script under the office's Policy 7.4. The empty slot is open and labeled, with a placeholder where the summary belongs.
Key points at the top · Policy 7.4 · Missing
The agent back-translates the Spanish form and aligns it with the English paragraph by paragraph. In ¶7, a child whose parent says no "will not take part in the reading class", while the English keeps the child in the regular lesson. That is a difference in meaning, a penalty for saying no under 46.116(b)(8); ¶8 and ¶9 differ in wording only.
“su hijo no participará en la clase de lectura”
Additional costs to families, 46.116(c)(3), is below the agent's threshold. Jae-won sees what the rule asks, that the form says nothing, and that protocol §10 says there are no costs to families. She marks it does not apply, and her reason code goes on the audit trail.
The agent drafts the return from the findings Jae-won kept: three changes and one suggestion, each with its paragraph, its rule and what to do. Number of surveys under 46.116(b)(1), the assent script's key points under Policy 7.4, Spanish ¶7 under 46.116(b)(8), and a suggestion to bring the reading level down. Jae-won can edit or remove any change with a reason, then sends it for a check.
Four returns wait for a check, with 2026-0418 at the top. Daniel sees each change with its evidence, then checks and sends it, or returns it to Jae-won with a reason. The analyst who prepared a return cannot check it.
Dr. Castellano sees the three changes, checked by Daniel Oyelaran on October 21. She asks how to fix each one; the answer cites the passage, the rule and, for ¶7, the back-translation, and offers wording she can use in version 2. She takes it and resubmits, and the agents re-check all four forms.
“Si usted dice que no, su hijo seguirá en la clase de lectura normal con sus compañeros.”
Version 2 reaches Dr. Osei as one brief: the request, Jae-won's recommendation, the criteria with their evidence, and only the paragraphs that changed, each tied to the finding it closes. Nothing else changed. He approves under expedited category 7 and 46.404, one parent's permission and the assent of each child, with a reason code.
One study, seen by the five people who carry it through IRB protocol review: what their day looked like, and what it looks like now.
Seven agents read every submission, tie each consent element to its paragraph and draft the return; the analyst completes it and a reviewer decides.
Checks each submission against the requirement set for its requested path (exempt, expedited or full board) and its populations, and reads the training records, the district approval and the data security plan.
Maps each form (adult consent, parental permission, child assent, exempt information sheet) to its required elements, checks that key information comes first, and cites the paragraph and the rule.
Compares facts across the protocol and every form: procedures, sessions, surveys and recordings, risks, records shared, retention, compensation and withdrawal.
Back-translates translated forms and flags differences in meaning, measures reading level by paragraph, and suggests plain-language rewrites.
Drafts the return to the investigator from the findings the analyst kept, in plain words, each with its citation, and drafts determination letters from the reviewer's decision.
On resubmission, compares every document with the last version and ties each change to the finding it closes.
Builds sealed, redacted exports of the audit trail and cited sources for an accreditation site visit, a federal inspection or a partner district.
Complete, check, determine. The agents propose; a named person decides.
The analyst can ask the assistant about any protocol, element or rule, or tell it what to check. Every answer opens its sources.
Why is 2026-0418 going back to the investigator?
Three changes and one suggestion, each cited: surveys, 3 in the protocol and 2 on the permission form; the assent script has no key points at the top; Spanish ¶7 reads as missing class when a parent opts out; and a suggestion, reading grade 12.4 against a target of 8. You decide what goes back; Daniel Oyelaran checks it.
Which submissions could miss the 3-day clock?
Two are on day 3 today and were escalated to Margaret Lindahl at 8:00 a.m.: 2026-0388 and A-26-1186, which is missing one district approval. Five more are on day 2. The queue is filtered to all analysts, ready for pre-review, longest wait first.
Can 2026-0427 be exempt?
Probably not under (d)(2). It surveys students in grades 9–12, and the (d)(2) exemption does not apply to surveys or interviews of children. It fits expedited category 7. Jae-won Park recommended expedited yesterday; Dr. Kwame Osei determines it. I never decide exemption.
Build the accreditation pack for this term
Every determination from July 1 to October 21, 612 in all, 233 of them exempt, each with its category, criteria, cited evidence, the analyst's recommendation and the reviewer's reason. Participant information is redacted, and the export is sealed and logged.
13 screens from the working solution, on its sample data. Pick one to see it large.
What each form must contain, how long each step may take and how sure an agent must be are settings, versioned on the audit trail. Rule changes need the HRPP director's approval.
| Setting | Default | Choose from |
|---|---|---|
| Requirement set · parental permission | 14 elements, rules v14 | Each element Required · Required when it applies · Recommended; adult consent 12, child assent 7, exempt information sheet 6 |
| Pre-review within | 3 business days of a complete submission | Escalate to the HRPP director on day 3, with the reason it is waiting |
| Investigator response | 30 days | Reminder at the end; withdrawn at twice the time |
| Designated reviewer determination | 5 business days | After pre-review is complete; full-board agenda closes 10 days before each meeting |
| Reading level | Grade 8 for parental permission and adult consent; grade 4 for child assent in grades 3–5 | A guideline: a miss is a suggestion, not a required change |
| Agent confidence threshold | 0.80 for the Consent agent | Per agent: 0.80 · 0.85 · 0.90 · 0.95 · 1.00; below it, a person decides |
| Reason codes for an override | 6 codes | Not applicable to this study · Met elsewhere in the form · Agent misread the paragraph · Not a real difference · Board practice · Other (note required) |
| Keep IRB records | 6 years after study closure | 3 years · 6 years · 10 years (45 CFR 46.115(b) requires at least 3) |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
Sources: 45 CFR 46, the revised Common Rule, including Subpart D (HHS Office for Human Research Protections) · 63 FR 60364, expedited review categories · 34 CFR 99.31(a)(6), FERPA · 20 USC 1232h, PPRA · a university IRB's published turnaround targets (initial pre-review screening within 2 weeks; 5 days for expedited or exempt review) · Journal of Clinical and Translational Science (2024): 5,239 US consent forms, average Flesch-Kincaid grade 10.99, 91% above grade 8; the 8th-grade level is a common IRB guideline, not a federal rule · "target" = design goal, measured in the live solution · "demo" = seen in the working solution, on its sample data · "typical" = published figures · "estimated" = our arithmetic. Halden State University, Marlow Bay Unified School District and the people named on this page are fictional: characters and sample data in the working solution.
IRB Protocol Pre-Review is the staff step of IRB protocol review, done with agents: IRB staff screen protocols, parental permission and consent forms for Common Rule elements before a reviewer sees them. Seven agents check completeness for the review path, tie every required consent element to its paragraph, compare the protocol with every form, read translated forms back, measure reading level, draft a cited return, compare each resubmission with the last version and build a sealed audit pack. The analyst completes the pre-review, a second analyst checks every return, and a designated reviewer, the chair or the board makes every determination.
That every document the requested path needs is attached; that each form carries the required elements of 45 CFR 46.116, Subpart D for children, FERPA and PPRA where they apply, and the office's own policies; that the forms say what the protocol says; that translated forms mean what the English means; and that families can read them. In the working solution, every one of those results opens the passage it rests on.
Under the revised Common Rule, 46.116(b) lists nine basic elements, including the purpose and procedures, risks, benefits, alternatives, confidentiality, whom to contact and that taking part is voluntary; 46.116(c) adds elements when appropriate, such as costs to participants. Key information must come first (46.116(a)(5)(i)). For children, Subpart D adds parental permission and the child's assent. The office sets which elements each form type requires, versioned.
Not under exemption (d)(2): 46.104(b)(3) says it does not apply to surveys or interviews with children. When a submission asks for that exemption, the Completeness agent flags it with the rule cited, the analyst recommends a path, and a designated reviewer decides. Agents never decide exemption.
It depends on the review path and the office. One university IRB publishes a target of 2 weeks for the initial pre-review screening and 5 days for an expedited or exempt review. The design goal for IRB Protocol Pre-Review is a cited pre-review back with the investigator on day 1 of a 3-day clock, measured in the live solution.
Yes. The Language agent back-translates each translated form, aligns it with the approved English paragraph by paragraph, and separates a difference in meaning from a difference in wording, naming the rule a meaning change touches. A bilingual reviewer confirms a translation finding when the analyst asks.
Yes. Agents read, compare and draft; they never determine exemption, expedited eligibility or approval, send anything to an investigator or edit a submitted document. The analyst can accept, edit or dismiss any finding with a reason code, a second analyst checks every return, and a designated reviewer, the chair or the board determines every study, with the reason on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents: your requirement sets, policies, form templates and a sample of past submissions. It goes live once every quality gate has passed. We will walk you through it on your own protocols first.
Five high-fit calls in minutes, each limit and cost share cited to the sponsor's own text.
1-min demoGrant Proposal ReviewEvery proposal reviewed inside the five-day window, every value tied to its clause.
1-min demoSubrecipient Risk Assessment & MonitoringRisk scored before signing and every invoice line checked before the university pays.
1-min demoWe'll run IRB Protocol Pre-Review on a sample of your own submissions and consent forms.